Improve GPU model classification and searchable hardware filters

[ziadmoubayed]

Summary

Replace broad GPU-prefix guesses with an explicit, sourced hardware catalog so GB200, GB300, B200 and B300 can be found and compared across providers.

• Correct GCP A4X / A4X Max and AWS P6 / P6e mappings; distinguish RTX PRO 4500/6000, A10/A10G, A100 memory variants, AMD GPUs and non-GPU accelerators.
• Search by hardware model, architecture, provider family and memory; select an entire model or individual families. Sticky, removable chips show exact hardware and retain selections when changing providers or searches.
• Keep provider-confirmed AWS P6e-GB200/P6e-GB300 and Azure ND GB200/GB300 selectable when regional coverage is incomplete. Show a confirmed provider badge and specification links.
• Separate offering confirmation from verified locations. Proximity results, provider comparisons and Markdown exports flag incomplete regional coverage, prevent misleading closest-provider conclusions, and never invent coordinates.
• Preserve existing family URLs and fail closed for unknown or removed selections.
• Validate catalog coverage after live provider refresh, retaining refreshed artifacts for review before classification checks.
• Refresh AWS data for Trainium2/Trn2u in Hyderabad and G7 in Frankfurt and Stockholm.

Verification

• 30 Python updater tests and 25 Node map/filter tests pass.
• Browser-verified model/family search and selection, persistent chips, URL reload, proximity results, AWS-only GB300 selection, incomplete-coverage comparison and Markdown export.
• Live-refresh validation passes: https://github.com/ziadmoubayed/cloud-accelerator-map/actions/runs/37009570067
• AWS GB300 preview: https://codex-gpu-filter-classificat.cloud-accelerator-map.pages.dev/?provider=aws&gpus=P6e-GB300

Official specifications

• https://docs.cloud.google.com/compute/docs/gpus
• https://aws.amazon.com/ec2/instance-types/p6/
• https://aws.amazon.com/about-aws/whats-new/2026/06/aws-parallel-computing-service/
• https://docs.aws.amazon.com/ec2/latest/instancetypes/ac.html
• https://learn.microsoft.com/en-us/azure/virtual-machines/sizes/gpu-accelerated/nd-family

AWS explicitly documents P6e-GB300. Its public regional dataset still omits this family. The Capacity Blocks availability table lists a Dallas Local Zone for P6e-GB200; that location is not assumed for GB300: https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/ec2-capacity-blocks.html

[cloudflare-workers-and-pages]

Deploying cloud-accelerator-map with    Cloudflare Pages

Latest commit:	cc96484
Status:	✅  Deploy successful!
Preview URL:	https://85c3612d.cloud-accelerator-map.pages.dev
Branch Preview URL:	https://codex-gpu-filter-classificat.cloud-accelerator-map.pages.dev

View logs